In a telecommunications system it is often useful to detect the language associated with a particular call. This can lead to several applications, for example, the automatic routing in a multi-lingual call center to the adequate operator or automatic speech recognition (ASR) queue, or for surveillance purposes as a first stage in the search of suspects that speak in a specific language. State of the art automatic language identification systems operate using pulse code modulation (PCM) and require the bit-stream to be decoded, a feature extraction representation to be computed, and the main recognition algorithm to be run.